ARRC's Polarimetric X-band Radar

					Examining the Impact of Spatial and Temporal
 Resolutions of Phased-Array Radar on EnKF
Analysis of Convective Storms Using OSSEs -
        Modeling Observation Errors



Yasuko Umemoto1, 2, Ting Lei3, Tian-You Yu1, 2 and Ming Xue3, 4
1Atmospheric Radar Research Center, University of Oklahoma
2School of Electrical and Computer Engineering, University of Oklahoma
3Center for Analysis and Prediction of Storms, University of Oklahoma
4School of Meteorology, University of Oklahoma
           Observing System Simulation
              Experiments (OSSEs)

•   The National Weather Radar Test Bed S-band Phased Array Radar (PAR)
    - Adaptively scan multiple regions using electric beam steering
    - Potential to increase warning lead time of severe storms

•   Assimilate the PAR data into the ARPS using EnKF
    - Use proper beam pattern and range weighting function to assimilate
    radial by radial observation
    - Extended OSSEs to examine additional capabilities of the PAR in more
    realistic settings
    - Examine the impact of data assimilation with various scanning strategies




                                                                          2
                  Observational Error Model
Z (ARPS output)    Estimated Error of     Simulated Z data            Calculate Z and Vr from ARPS
                           Z

                                                                                      Add random noise

                                                                                                Error  SD * random
                                                                      Simulated observation

                                                                          
  Vr (ARPS         Estimated Error of       Simulated Vr
   output)                Vr                   data               Estimated Standard Deviation
                                                                  - Distance from the Radar
                                                                  - Reflectivity (SNR)
                                                                  - Scanning strategy (Number of pulses)
                                                                                            1
                                    M 1
                                               M  m 
                                                             1
                                                            1                           
                                                                                          2
                                                                                          
                                    
                  SD[Z ln ]  Z ln  m (M 1)
                                                M  SNR 
                                                    2
                                                       1
                                                        
                                                        
                                                               mTs  1 SNR  m,0 
                                                                                  1
                                                                                          
                                                                                          
                                                                                       
                                                                                          
                                                                                                                     1
                                                                                        1 2            1  2
                                                                                                                
                                             
                                                             M 1
                                                                                1 1
                                                            
                                            1
                  SD[V ]   32 Ts  Ts  M 1   Ts    2 (mT)(M  m ) 
                             2  2 2 2         2   2
                                                                                      2         (2Ts ) 1 
                                                                                                 1             
                                                        m (M 1)
                                                                                M SNR     M SNR         M  
                                                                                                                
             


                                  
                  mTs   exp 8 v mTs / 
                                                    2
                                                        

                                                                                                               3
Observational Error Model




                            4
                  Experimental Design



•   Perfect prediction model
•   64X64X20 km3 with 43 vertical levels
•   1 km horizontal resolution
•   1977 May 20th Del City OK super cell storm
•   Same model and configuration for the truth and ensemble forecast
•   The initial ensemble forecast starts at 20 min of the model time
•   Positive Z (dBZ) and Vr data where Z > 5 dBZ are assimilated.




                                                                       5
                   Experimental Design
•   RAR (2o beam) or WSR-88D (1o beam) were assimilated
•   Observation with 14 elevation angles
•   1 or 0.5o azimuthal and elevation Over-Sampling (OS) with PAR
•   Rapid or slow update scanning (2 v.s. 5 min.)
•   Distant or nearby storms ((-100,0) v.s. (0,0) of model grid)




                                           100 km
                                                                    6
                Slow Update (5min)
Distant storm    Ensemble mean forecast and analysis of RMS




                                     PAR-OS shows as good performance as 88D



Nearby storm




                                           Improvement due to OS is not obvious




                                                                             7
                Fast update (2min)
Distant storm                      PAR-OS and 88D with fewer sampling




Nearby storm




                            PAR-OS shows better result than 88D

                                                             8
      Fast updates – Fewer samplings – Degradation of data accuracy

      PAR – adaptively scan by electric beam steering
          -- allows fast update of weather info without compromising data accuracy



Truth simulation   WSR-88D 5min   PAR 5min   PAR-OS(1o) 5min     PAR 2min        PAR-OS(1o) 2min
  w


                                                                             So no OS
                                                                             here?


  Z




                                                       Analysis of w and Z at t = 3600 s, 1km height




                                                                                        9
     Scan-strategy dependent/independent
                 error model
                Distant storm   Nearby storm



Scan-strategy
dependent




Scan-strategy
independent




                                               10
                Adaptive scan-strategy


Not Adaptive
14 elevations




Adaptive scan
Cover whole
storm



                                         11
                              Summary


•   The earlier OSSEs are extended to examine additional scanning strategies
    possible with PAR, and using more realistic settings.
•   Agreeing with earlier results, azimuthal and elevation oversampling with 1
    or 0.5 degree increments and rapid updates improve the EnKF analysis.
•   For these experiments, observation errors that are spatially
    inhomogeneous and scanning strategy dependent are applied.
•   By properly modeling the expected error in the observations, the OSSE
    results are more robust.
•   An optimal combination of the spatial and temporal resolutions and data
    accuracy is sought through the EnKF OSSEs.




                                                                         12
List of Experiments




                      13

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:3
posted:6/27/2011
language:English
pages:13